Repository navigation
chore(a2ml): move the chora-registry off A2ML, within A1's guardrails - #115
Conversation
…ecs, move the readers, delete 114 of 117 files A2ML is retired (D99, D269c, D312, D313). This follows hypatia#1113: translate only what a mapping spec in standards 1-formats/deed/mappings/ covers, record everything else in docs/history/A2ML-REMOVED.adoc (path, size, sha256, blob id, what it held, where it went), then delete. New deeds (each passes deed_lint at standards b2aaeb0c): - gv-clade-index_chora.deed, the repo deed this repo lacked. CLADE.a2ml -> (identity) (clade) (forges) (lineage) per clade-to-repo-deed.adoc; AGENTIC/NEUROSYM/PLAYBOOK -> (agentic) (neurosym) (playbook) per agentic-neurosym-playbook-to-repo-deed.adoc. :repo-uuid is the v8 id chora-registry already holds for this repo (a65b74d8-…, :minted-in gv-clade-index), written as #uuid; nothing was minted or reminted. :beholding-chora is the estate chora the seed records. - .machine_readable/descriptiles/STATE.deed: CLADE's [status] (A7 J15, as the clade spec now directs) and STATE.a2ml's maturity (state-v1-decision option B). Passes --state --strict against the chora deed. - 0-AI-MANIFEST.deed, pointer-only (A7 Q15). Not translated: META and ECOSYSTEM use section shapes the meta-ecosystem table does not list (P-2a), and ECOSYSTEM's related projects use relationships outside the closed set. ANCHOR has no spec in mappings/ yet (A7 6.4 is unratified) and held template defaults. The 81 directory manifests (0.N-AI-MANIFEST.a2ml) are scaffold boilerplate with no repo-specific fact, ledgered as one class. The contractiles, integrations, policies, agent-instruction templates, ENSAID_CONFIG, clade-hygiene and the docs/governance and docs/templates copies have no mapping spec (Q7). Retained (registry data with live readers, each waiting for its own mapping spec): verisim/seed/clades.a2ml, verisim/seed/forges.a2ml, verisim/config.a2ml. The seed deed, sync/, check-registry.sh and its test are untouched. Readers moved: Justfile (info, init, validate-rsr, validate-state, state-phase; state-touch removed), rhodibot.yml (wants 0-AI-MANIFEST.deed, warns on a retired A2ML manifest), MUST.contractile, bust.ncl, contractile.just, CODEOWNERS, worker/src/validate.js (comment), ai.txt, the agent rule files and AI-CONVENTIONS, the PR template, CONTRIBUTING, README and the docs that pointed at deleted files. Checks: scripts/check-registry.sh PASS (and PASS with GV_BASE_REPOS_JSON = main's repos.json: all 327 repos kept, ids unchanged); tests/check-registry-test.sh PASS; check-uuid-v8.sh --strict reports the same findings as main (none new); validate-a2ml.sh 0 errors; just fmt-check and just lint clean; yq parses the changed workflow. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC
|
Important Review skippedToo many files! This PR contains 174 files, which is 74 over the limit of 100. To get a review, reduce the PR to 100 files or fewer by splitting it into smaller PRs or changing its base branch. Upgrade to a paid plan to raise the limit. This review couldn't start because sufficient usage credits or metered capacity aren't available. Add credits or update usage-based reviews in the billing tab, then retry.
|
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c0f477f4d7
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| for f in AI.a2ml 0-AI-MANIFEST.a2ml; do | ||
| [ -f "$f" ] && warn "retired A2ML manifest present: $f (the front door is 0-AI-MANIFEST.deed)" |
There was a problem hiding this comment.
Scan all retired A2ML manifests
The Rhodibot workflow only checks the two root filenames, so reintroducing any of the retired nested manifests—such as .machine_readable/CLADE.a2ml or a deleted 0.N-AI-MANIFEST.a2ml—still produces a clean canary result. Since this change establishes that only the three legacy registry-data .a2ml files remain allowed, scan repository-wide and exclude that explicit allowlist rather than checking just these two names.
Useful? React with 👍 / 👎.
| operations only. `ROADMAP.adoc` tracks phase status (the A2ML `[route-to-mvp]` | ||
| it replaced was retired 2026-11-14), and this runbook tells you when and how | ||
| to update it. Filed as #31. |
There was a problem hiding this comment.
Update all phase bookkeeping instructions to ROADMAP
When a rollout phase closes, this newly designated status source conflicts with the operating rule immediately below, which still requires updating [route-to-mvp] in the now-deleted STATE.a2ml; later phase gates repeat the same obsolete instruction. A maintainer following this authoritative runbook can therefore skip the real ROADMAP.adoc update or attempt to edit a nonexistent file, so the phase bookkeeping and gate instructions need to be migrated together.
Useful? React with 👍 / 👎.
| # STATE FILES (.machine_readable/descriptiles/ ONLY) | ||
| # Repo metadata is deed (gv-clade-index_chora.deed, descriptiles/STATE.deed). | ||
| # A2ML is retired: never create a .a2ml file. | ||
| # The .machine_readable/ directory is the single source of truth. |
There was a problem hiding this comment.
Remove the conflicting single-source instruction
The updated rule tells agents that repository metadata lives in the root gv-clade-index_chora.deed, but the next unchanged line still declares .machine_readable/ to be the single source of truth. In sessions driven by this tool-specific configuration, those instructions permit an agent to treat the root deed as non-authoritative even though the new front door and validation require it, so the authority statement should explicitly include both deed locations.
Useful? React with 👍 / 👎.
| Repo metadata is deed: `gv-clade-index_chora.deed` at the root, and descriptile | ||
| deeds (`STATE.deed` and the like) in `.machine_readable/descriptiles/` ONLY. | ||
| A2ML is retired: NEVER create a `.a2ml` file (`docs/history/A2ML-REMOVED.adoc`). |
There was a problem hiding this comment.
Migrate the K9 setup generator before banning A2ML
With this rule in force, the documented Hunt-level setup component still enables checkpoint_files by default and writes STATE.a2ml, ECOSYSTEM.a2ml, and META.a2ml in .machine_readable/svc/k9/examples/setup-repo.k9.ncl:51,123-127. Anyone copying that component as instructed by the K9 README generates files this commit declares retired, so the generator must emit the replacement deeds or disable that feature before the blanket prohibition is published.
Useful? React with 👍 / 👎.
| # Show the current lifecycle phase from STATE.deed. (The old state-touch recipe, | ||
| # which bumped a date in STATE.a2ml, is gone: STATE.deed's :affirmed date records | ||
| # a review, so it is changed by hand, never by a timestamp bump.) |
There was a problem hiding this comment.
Remove the documented state-touch command
Removing state-touch leaves docs/RSR_OUTLINE.adoc:122 advertising just state-touch as the supported STATE-management command. A user following that repository guide now gets an unknown-recipe error, so the documentation needs to be changed to the new manual :affirmed review process (or a replacement recipe must remain available).
Useful? React with 👍 / 👎.
…117) ## Summary On 2026-11-14 GitHub renamed `hyperpolymath/gv-clade-index` to `hyperpolymath/chora-registry`. This PR carries out the follow-up that #115 named. No id changes. The repo keeps `#uuid"a65b74d8-351a-89d9-a38d-aea470863b03"` (rule U1). That id keeps its `gv-clade-index` minting domain for good (rule V9). ## Changes - **The repo's own deed.** `gv-clade-index_chora.deed` is renamed `chora-registry_chora.deed`. - `:canonical-name`, `:prefixed-name` and `(forges :github)` take the new name. - `(lineage)` gains `:previous-names ("gv-clade-index")`. - `:repo-uuid` is unchanged, and the file lints OK. - **The registry's record of itself.** In the seed, its entry becomes `:name "chora-registry"`. - It gains `(former :generation 1 :path "hyperpolymath/gv-clade-index" :until "2026-11-14")`. - `uuid` and `uuid_v8` are unchanged, because the birth path is the same. - `archive.json` gains the move record. - `repos.json` and `index.json` now serve the new `name`, `github`, `prefixed` and `description`. - **Gate fix.** Section 7 of `scripts/check-registry.sh` (never removed) now follows a repo by its id, not by its name. - Main's gate fails this rename. It reports "removed from the seed: gv-clade-index", even though the id is still registered. I ran main's gate against this tree to confirm. - A real removal still fails, and so does an id change: planted cases 17 and 18. - A rename now passes: new planted case 24. - **Files that name the deed** now use the new file name: the AI rule files, `0-AI-MANIFEST.deed`, README, QUICKSTART, `www/.well-known/ai.txt` and `docs/practice`. - The `Justfile` `project` and `REPO` values also take the new name. - So do the `wrangler.toml` `DATA_SOURCE_URL` and the worker's `source` link. - **Docs.** REPO-ONBOARDING no longer calls a clade placement an "optional finding aid". A placement is an identity fact (A7 W10). **Deliberately not changed:** - The `gv-clade-index` minting domain in `sync/` and in the seed's `:minted-in` (V9). - The cache directory names. - The served `registry: 'gv-clade-index'` field in `worker/src/index.js`. That is an API value; changing it is the owner's call. - Historical docs: ADRs, reports, CHANGELOG and `docs/history`. ## Testing - `bash scripts/check-registry.sh` passes. With `GV_BASE_REPOS_JSON` set to main's `repos.json`, all 327 base repos are still registered, followed by id, with ids unchanged. - `bash tests/check-registry-test.sh` passes. The 19 planted failures fire, and the rename case passes. - `shellcheck -S warning` is clean. - `deed_lint` reports OK on `chora-registry_chora.deed`. This PR stays a draft for the owner. The repo has no stamped proven-lock-gate, so I do not self-merge it. --- _Generated by [Claude Code](https://claude.ai/code/session_0136P7ZoUeC49fDEvSwDNvuS)_ Co-authored-by: Claude <noreply@anthropic.com>
## Summary #115 merged with five Codex P2 findings still open. This PR fixes them. A1, who owns this repo, asked for it as a PR separate from the rename (#117). It is cut from main 57607d3, which already includes #117. ## Changes 1. **`.machine_readable/svc/k9/examples/setup-repo.k9.ncl`** - `checkpoint_files` now defaults to `false`, with a comment pointing to the root `<repo>_chora.deed` and `.machine_readable/descriptiles/STATE.deed`. - The `create-checkpoint-files` recipe no longer writes STATE/ECOSYSTEM/META `.a2ml` files. It prints a pointer to the deed and exits 1. - No new deed form is invented. 2. **`docs/governance/RELEASE-RUNBOOK.adoc`**: the bookkeeping rule and the Phase 2/3/4/6 gates now update the phase Status in the root `ROADMAP.adoc`. The VPS fact cites `docs/SPEC-clade-verisim-portal.adoc` Layer 3. No `STATE.a2ml` references remain. 3. **`.github/workflows/rhodibot.yml`** - The workflow now scans the whole repo, `9-archive/` included, and fails on any `.a2ml` except the three allowlisted files: `verisim/config.a2ml`, `verisim/seed/clades.a2ml` and `verisim/seed/forges.a2ml`. - No `${{ }}` in `run:`, and the `uses:` lines are unchanged. 4. **`.machine_readable/ai/.clinerules`, line 16 only**: identity is the root `chora-registry_chora.deed`; descriptive deeds live in `.machine_readable/descriptiles/`. 5. **`docs/RSR_OUTLINE.adoc:122`**: `just state-touch` is replaced by the manual `:affirmed` review. The Justfile is not edited. ## Verification - **Rhodibot scan**, step extracted with yq and run with bash: - current tree: clean, rc=0; - planted `9-archive/old/STATE.a2ml` plus `docs/stray.a2ml`: rc=1; - only the `9-archive/` plant: rc=1; - planted root `0-AI-MANIFEST.a2ml` plus the near-miss `verisim/seed/x/clades.a2ml`: both flagged, rc=1; - plants removed: clean again. - `scripts/check-registry.sh`: PASS, 327 repos. Also PASS against main's `repos.json`: no id changed. - `tests/check-registry-test.sh`: PASS, all 19 planted failures fire. - `just fmt-check`, `just lint` and `just validate-state`: OK. - `nickel typecheck` and `nickel export` on the k9 component: OK. - A1's guardrail files are untouched, no `.a2ml` is created, and no id is changed. ## Departures seen, not fixed - `docs/RSR_OUTLINE.adoc` is still a stale template copy: it recommends A2ML files, says to use Deno and not Bun, lists ReScript as tier 1, and carries Palimpsest badges (licence material, flag only). - `.clinerules` lines 27–30 reverse the current policy ("TypeScript → ReScript", "bun → Deno"), and line 3 points at a missing `.md`. - The `add-license` recipe in `setup-repo.k9.ncl` fetches PMPL-1.0. That is licence drift, flag only. - `dogfood-gate.yml` interpolates `${{ steps.detect.outputs.* }}` into `run:`. --- _Generated by [Claude Code](https://claude.ai/code/session_0136P7ZoBy5xGnGDLk7wQ8jC)_ Co-authored-by: Claude <noreply@anthropic.com>



Summary
The owner approved this on 2026-11-14, and A1, who owns this repo, agreed. It follows hyperpolymath/hypatia#1113 and deletes 114 of the 117
.a2mlfiles.docs/history/A2ML-REMOVED.adocrecords each one (path, size, sha256, blob id and disposition), plus the files kept and the questions still open.Changes
gv-clade-index_chora.deed. The registry had no repo deed, so this is a judgement call for A1 to confirm.CLADE.a2ml(identity, clade gv with secondary db, forges, lineage) and from AGENTIC, NEUROSYM and PLAYBOOK, following their mapping specs.:repo-uuid #uuid"a65b74d8-351a-89d9-a38d-aea470863b03"is the id the seed already holds (:minted-in gv-clade-index). Nothing was minted.chora-registry_chora.deedwhen GitHub renames the repo; the id stays the same..machine_readable/descriptiles/STATE.deed: the CLADE status and STATE maturityalpha. It passes--state --strict.0-AI-MANIFEST.deed: pointer-only.ENSAID_CONFIG,ai/AI.a2ml, and the agent_instructions, contractiles, integrations and policies files;clade-hygiene.a2ml;.adocversions stay;verisim/seed/clades.a2ml, which builds the servedclades.json;verisim/seed/forges.a2ml, covered by the seed lint and still waiting for its own conversion;verisim/config.a2ml, cited bywrangler.tomland the release runbook.check-registry.shand its test.rhodibot.ymlchecks for0-AI-MANIFEST.deed;MUST.contractile,bust.ncl,contractile.justand CODEOWNERS;worker/src/validate.js;Verification
scripts/check-registry.sh: PASS. WithGV_BASE_REPOS_JSONset to main'srepos.jsonit also passes: all 327 repos are present and no id changed.tests/check-registry-test.sh: PASS.check-uuid-v8.sh --strict: no new findings (the same 702 as main).validate-a2ml.sh,just fmt-check,just lintand the yq workflow parse all pass.Open questions
These are in the ledger:
depends-on,feeds-into,sibling);mappings/;Generated by Claude Code